The Agent Skills Directory

[SAFE]: The skill functions as a documentation assistant, focusing on maintaining a glossary and recording architectural decisions. It reads project files and writes Markdown documentation, which are standard and safe operations for this utility.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection, as it processes content from untrusted project files and source code to inform its modeling tasks.
Ingestion points: Reads CONTEXT.md, CONTEXT-MAP.md, and project source code in the src/ directory.
Boundary markers: None; the skill does not use specific markers to delimit untrusted data or instruct the agent to ignore instructions within that data.
Capability inventory: The skill is capable of writing to CONTEXT.md and creating new Markdown files in docs/adr/.
Sanitization: There is no explicit sanitization or validation of the data read from project files before it is processed or written back to documentation.
Mitigation: To reduce the risk of indirect prompt injection, external content should be wrapped in delimiters with explicit instructions for the agent to ignore embedded commands.

domain-modeling